We’ve got time to develop some comprehensive benchmarks of other proposed “real-time Linux” solutions and I’ve started reading up on published benchmarks. We had the advantage of a large number of customers with real applications to assist us in figuring out what to benchmark and the standard tests with RTLinux and RTLightning (FSMLabs version of RTLinux optimized for enterprise) are reasonably solid. So it is fascinating to read other people’s benchmarks and try to figure out what applications they had in mind. One thing that stands out is the brevity of testing. For example in Andrew Webber’s 2002 Linux journal article, there is an interesting section title
Largest interrupt latency time measured over 3.5 million samples.
Well, if the interrupt frequency is 1megahertz, then that is under one hour. The headline
Largest interrupt latency time measured over almost one hour.
would be less impressive sounding.
(RTLinux is a trademark of WindRiver Systems, RTLightning is a trademark of FSMLabs).